A journey from phosphotyrosine to phosphohistidine and beyond

Summary:

Protein phosphorylation is a reversible post-translational modification. Nine out of the 20 natural amino acids in proteins can be phosphorylated, but most of what we know about the roles of protein phosphorylation has come from studies of serine, threonine and tyrosine phosphorylation. Much less is understood about the phosphorylation of histidine, lysine, arginine, cysteine, aspartate and glutamate, so-called non-canonical phosphorylations. Phosphohistidine (pHis) was discovered 60 years ago as a mitochondrial enzyme intermediate, since then evidence for the existence of histidine kinases and phosphohistidine phosphatases has emerged, together with examples where protein function is regulated by reversible histidine phosphorylation. pHis is chemically unstable and has thus been challenging to study. However, the recent development of tools for studying pHis has accelerated our understanding of the multifaceted functions of histidine phosphorylation, revealing a large number of proteins that are phosphorylated on histidine and implicating pHis in a wide range of cellular processes.

In this perspective, Hunter reviews the history of protein-histidine phosphorylation, the current status of the field, and the prospects for further advances in understanding the functions of histidine phosphorylation in regulating cellular processes.

Since the first issue of Molecular Cell was published in December 1997, the amazing genomic and proteomic complexity of protein phosphorylation has come to the fore. A complete catalogue of the human kinome was reported in 2002, confirming that protein kinases are one of the largest protein families with over 500 protein kinase genes (Manning et al., 2002). A 2004 compilation of human protein-tyrosine phosphatases (PTPs) recorded over 100 PTP genes, and a more recent catalogue of the entire protein phosphatase universe, which together counteract the activities of the huge number of protein kinases, revealed 189 human genes encoding protein phosphatase catalytic subunits (Alonso et al., 2004). In toto, the number of protein kinase and phosphatase genes combined with those of their regulatory subunits accounts for nearly 2.5% of all human genes.

Twenty-five years ago, we had already surmised that the total number of protein phosphorylation sites would be large, based primarily on 2D gel analysis of ³²P-labeled cellular proteins, but it was not until 2004 that the first systematic mass spectrometry (MS)-based phosphoproteomic survey was published (Beausoleil et al., 2004). Since then phosphoproteomic technology has improved by leaps and bounds, and, routinely, thousands of pSer, pThr, and pTyr phosphosites can now be identified from cultured cells or tissues within a few hours. HeLa cells reportedly possess as many as 50,000 different phosphosites (Sharma et al., 2014), and, as of the end of January 2022, the PhosphoSitePlus^™ database lists ~240,000 unique human phosphosites. Even though many of these sites have been identified only once or twice, this is an extraordinarily large number given that the human genome has only ~20,000 protein coding genes.

Over the past 25 years, the primary emphasis of protein phosphorylation studies has been to understand how canonical phosphorylation of the three hydroxy amino acids in proteins, serine (Ser), threonine (Thr) and tyrosine (Tyr), regulates cell and organismic functions. However, what is often overlooked is that six of the non-hydroxy amino acids can be phosphorylated in proteins – histidine (His), arginine, lysine, aspartate, glutamate and cysteine. Such non-canonical phosphorylations have begun to emerge as an important new area that will undoubtedly add further complexity to the protein phosphorylation landscape. What we have learned about the phosphorylation of histidine over the past few years and what the future holds for the histidine phosphorylation field will be the main focus of this perspective.

Getting from tyrosine phosphorylation to histidine phosphorylation

The possibility that tyrosine might be phosphorylated in proteins was first envisaged by Phoebus Levene in the 1930’s, through his efforts to characterize the phosphorylated hydroxyamino acids in phosvitin/vitellin and other abundant naturally-occurring phosphoproteins. Since it was unclear which of the three hydroxyamino acids might be phosphorylated in vitellin, Levene’s group chemically synthesized phosphoserine (pSer) (Lipmann and Levene, 1932) and phosphotyrosine (pTyr) (Levene and Schormüller, 1933), and used them as markers to identify which phosphoamino acids were present in acid hydrolysates of vitellin. Although vitellin proved to contain only phosphoserine, the concept of protein-tyrosine phosphorylation had been born. Nevertheless, it lay fallow for over 45 years until we discovered that polyomavirus middle T antigen was phosphorylated on tyrosine in an in vitro kinase reaction (Hunter et al., 1979). While this finding alone did not establish that proteins are phosphorylated on tyrosine in vivo, very soon thereafter, we showed that the Rous sarcoma virus (RSV) v-Src transforming protein exhibited intrinsic tyrosine kinase activity in vitro, and, more importantly, that v-Src protein expression led to a ten-fold increase in the levels of pTyr in proteins in RSV-transformed chicken embryo cells (Hunter and Sefton, 1980). The field of tyrosine phosphorylation had officially begun.

Over the past 40 years, the importance of tyrosine phosphorylation in cellular signaling has become abundantly clear, with its functions ranging from surface receptor signaling to cell cycle progression and neural transmission. Importantly, by serving as a new mode of intercellular communication, initially in single-celled eukaryotes, the evolution of tyrosine phosphorylation played a key role in the successful development of multicellular organisms. Early on, studies of tyrosine phosphorylation revealed new principles of phosphorylation-mediated signal transduction, highlighted by the discovery of modular pTyr-binding domains, such as the SH2 domain, which promote phosphorylation-induced protein-protein interactions that drive signal transduction pathways. Even though pTyr represents only 1% of all the phosphate attached to proteins in a mammalian cell, thousands of tyrosine phosphorylation sites have been identified; the occupancy of many of these sites changes in response to external stimuli, but the functions of only a minor fraction have been elucidated. Perhaps, the most important insight was that dysregulated tyrosine phosphorylation plays a driver role in many human diseases, particularly cancer, a discovery that has motivated the development of a suite of small molecule tyrosine kinase inhibitors (TKIs) for the treatment of cancer and inflammatory diseases. As of this writing, over 60 TKIs have been approved for clinical use, together with 11 approved protein drugs that antagonize signaling by receptor tyrosine kinases (Attwood et al., 2021).

Progress in deciphering the functions of tyrosine phosphorylation was relatively rapid, facilitated by the fact that pTyr, like pSer and pThr, is chemically stable. pTyr was initially detected in cell proteins by labeling cells with ³²P-orthophosphate, followed by partial acid hydrolysis of isolated proteins, which releases free phosphoester-containing amino acids, such as pTyr. However, the use of large quantities of ³²P is not for everyone, and, early on, a key technological development enabling more facile studies of tyrosine phosphorylation without the use of ³²P was the development of pTyr-specific antibodies for use as research tools. Amazingly, the first rabbit polyclonal anti-pTyr antibodies were reported within two years of the discovery of tyrosine phosphorylation (Ross et al., 1981)! Soon thereafter, mouse pTyr mAbs were described, and these became the workhorses of the field, greatly accelerating progress in unveiling the functions of tyrosine phosphorylation (Frackelton et al., 1983). Intriguingly, one of these first pTyr mAbs, 2G8, was found to recognize not only pTyr, but also phosphohistidine (pHis) in the ATP citrate lyase (ACLY) metabolic enzyme, where pHis serves as a phosphoenzyme intermediate (Frackelton et al., 1983). This implied that the mammalian immune system might be capable of generating pHis-specific antibodies if presented with the right antigen, and this has proved to be the case.

Histidine phosphorylation – past:

Phosphohistidine, first described in 1962, is one of the six non-canonical phosphoamino acids – pHis, pLys, pArg, pAsp, pGlu, and pCys - that have all been reported as protein modifications. pHis is unique in being the only phosphoamino acid that has two isoforms, 1-pHis and 3-pHis, with phosphate linked to either the N-1 or N-3 position on the histidine imidazole ring, respectively (Figure 1). Phosphohistidine was initially discovered as a posttranslational modification (PTM) by Paul Boyer, who reported that the mitochondrial enzyme succinyl CoA synthase (SCS) contains pHis (Boyer et al., 1962)(for a timeline of discoveries in the histidine phosphorylation field see Figure 2). This was later shown to be a 3-pHis residue in the SCSα subunit, whose formation is an essential step in the reaction catalyzed by SCS: succinyl CoA + Pi + NDP ↔ succinate + CoA + NTP. This reaction serves as a key step in the TCA cycle, generating a molecule of ATP or GTP, in a process known as “substrate level phosphorylation”. Later, pHis was discovered to act as a phosphoenzyme intermediate in several other metabolic enzymes, including ACLY (3-pHis), phosphoglycerate mutase (PGAM) (3-pHis), the nucleoside diphosphate kinases (NDPK) (1-pHis), and phospholipase D (3-pHis) (Table 1), as well as being the key active site conjugate in the histidine (His) kinase component of bacterial two-component signaling (TCS) systems (1- or 3-pHis) (for review see (Kalagiri and Hunter, 2021)). Bacteria use TCS systems to respond to external stimuli, such as amino acids, with the His kinase component transferring the pHis phosphate onto an Asp residue in a response regulator protein that then provides the signal output; this type of signaling system was largely lost during the evolution of multicellular eukaryotes.

Figure 1: Enzymatic phosphorylation and dephosphorylation of histidine residues in proteins.

Histidine protein kinases transfer the γ-phosphate from ATP onto the 1- or 3-nitrogren of the imidazole ring of histidine residues in proteins. The NME1/2 histidine kinases transfer the phosphate in a two-step process via a phosphoenzyme intermediate resulting in inversion of phosphate chirality (this step is not depicted in the figure). Phosphohistidine phosphatases (PHPs) hydrolyze the phosphate from the 1- or 3-position of the histidine imidazole ring releasing a free phosphate ion (P_i). In the case of the PGAM5 PHP, the P-N bond in the target pHis is attacked by the active site His of PGAM5, forming a transient pHis intermediate that is then hydrolyzed.

Figure 2: Timeline of key discoveries in the field of protein histidine phosphorylation.

Beginning with the report of covalent phosphorylation of proteins in 1906, and the identification of phosphoserine in proteins in 1932, the timeline shows some important milestones in histidine phosphorylation following the discovery of phosphohistidine (pHis) as a protein modification in 1962. Abbreviations: ACLY: ATP citrate lyase; KCa3.1: calcium-activated potassium channel 3.1; NDPK-A: nucleoside diphosphate kinase A; PHPT1: phosphohistidine phosphatase 1; P’ase: phosphatase; P’ation: phosphorylation; SCS: succinyl CoA synthase; TCS: two-component system.

Table 1.

Histidine kinases, pHis phosphatases and their substrates

Histidine kinases	Substrates
NME1	ACLY, Gβ1
NME2	Gβ1, KCa3.1, TRPV5
Histone H4 kinase	Histone H4
Bacterial TCS His kinases (e.g. CheA, NRii, EnvZ)	CheY, NRI, OmpR
pHis phosphatases
PHPT1	pACLY, pGβ1, pKCa3.1, pTRPV5
LHPP	?
PGAM5	pNME2
PP1, PP2A, PP2C	pHistone H4
pHis enzyme intermediates	Intermediate pHis isoform
6-phosphfructo-2-kinase bisphosphatase	3-pHis
Phosphoglycerate mutase 1	3-pHis
Acid phosphatase	3-pHis
STS-1, STS-2, PGAM5 phosphatases	3-pHis
Succinyl-CoA synthetase	3-pHis
Phospholipase D	3-pHis
Tyrosyl DNA phosphodiesterase 1	3-pHis
Glucose-6-phosphatase	3-pHis
NAMPT	1-pHis
NME1–9	1-pHis
Bacterial TCS His kinases	1-pHis or 3-pHis

If pHis only served as an enzyme intermediate, the study of histidine phosphorylation as a regulatory mechanism would probably have stagnated. However, because the free energy (ΔG°) of hydrolysis of the pHis phosphoramidate (P-N) bond is −12 to −14 kcal/mol, considerably higher than that of the phosphomonoester bonds in pTyr/pSer/pThr, the phosphate on pHis acts as an excellent leaving group, and can serve as a powerful phosphate donor. Indeed, the phosphate from a pHis enzyme intermediate can be transferred to a variety of acceptor amino acids, including Asp, Ser or His itself, in a second protein, as well as to small molecules, such as nucleoside diphosphates (Kalagiri and Hunter, 2021). Theoretically, then, a pHis-containing enzyme could act as a protein-histidine kinase by transferring its active site phosphate onto a target His. Consistent with His phosphorylation being a regulatory mechanism, subsequent studies uncovered proteins lacking enzyme activity that contain pHis, including histone H4, which is phosphorylated at His18 in its N-terminal tail or His75 in its core, the β subunit of a trimeric G protein, and a calcium-activated K⁺ channel, KCa3.1. For such proteins, site-specific His phosphorylation can in principle serve as a reversible post-translational modification that regulates protein function, analogous to Ser, Thr and Tyr phosphorylation.

In contrast to pTyr with its chemically stable phosphate ester P-O linkage, the P-N linkages in the 1- and 3-pHis isoforms are both heat and acid labile, due to the thermodynamic instability of the P-N bond (Kee and Muir, 2012). The chemical lability of pHis means that it cannot readily be detected or analyzed using the conventional methods that work so well for studying the chemically stable pSer, pThr and pTyr phosphoamino acids. As a result, identifying sites of His phosphorylation is challenging, and development of new methods was essential to detect and explore the biology of pHis. By analogy with the pTyr antibodies, which have proved so valuable, a possible approach for studying this unstable modification would be to generate pHis-specific antibodies that could be used to detect pHis in proteins under neutral conditions.

Piqued by the cross-reaction of the 2G8 anti-pTyr mAb with pHis, and the increasing evidence that His phosphorylation was going to be of broad biological significance, we decided to try to generate pHis-specific antibodies. In 1989, we set up a collaboration with Jonathan Reizer in Milt Saier’s group at UCSD, who were studying bacterial phosphotransferase systems used for sugar transport. Our plan was to use purified 3-pHis-IIA^Glc, a pHis-containing protein from the B. subtilis phosphoenolpyruvate:sugar phosphotransferase system (PTS) with an unusually stable 3-pHis phosphoenzyme intermediate (Anderson et al., 1992), as an immunogen to try to raise pHis antibodies. I persuaded a new graduate student, Martin Broome, that this might be an interesting side project for him to try in parallel with his main project to purify active Src tyrosine kinase. Using a routine rabbit immunization procedure, Martin immunized rabbits with 3-pHis-IIA^Glc in complete Freund’s adjuvant, and then boosted with 3-pHis-IIA^Glc in Freund’s incomplete adjuvant (see Figure S1 for excerpts from the relevant pages from his 1989/90 lab notebooks). To test whether the immune serum contained antibodies that recognized pHis, Martin used ELISA assays with purified 3-pHis-IIA^Glc, purified PGAM phosphorylated in vitro by addition of 2,3-diphosphoglycerate to generate 3-pHis11 PGAM, and synthetic polyhistidine phosphorylated chemically with phosphoramidate, which mostly generates 3-pHis. Disappointingly, no pHis antibodies were detected, even after a second boost, and we decided to abandon these efforts. At around the same time, Art Weiss’ and Lew Cantley’s groups had also tried to raise anti-pHis antibodies, also without success. We all concluded that the pHis phosphoramidate linkage in the immunizing pHis protein was too unstable to survive uptake, processing, and presentation as pHis-peptide/MHC complexes on the surface of antigen-presenting dendritic cells.

To be successful in generating pHis antibodies we knew what was needed were nonhydrolyzabe pHis analogues that could be used as haptens, but, while it was easy to design such pHis analogues with a stable phosphonate P-C bond in place of the P-N bond in the imidazole ring, I could not find a chemist willing to devise a synthesis for such analogues. While I continued to be interested in histidine phosphorylation, our efforts to develop pHis antibodies rested for the next 20 years (!), until 2010, when Jacques Mauger, a chemist at Sanofi, pointed out to me a recent paper by Tom Muir’s group at the Rockefeller University, in which they had devised facile syntheses of stable phosphonate analogues for both 1-pHis and 3-pHis using click chemistry with either Ru(II) or Cu(I) metal ions as catalysts to link two starting compounds (Kee et al., 2010). The syntheses generated ethyl ester-protected N-Boc derivatives of 1-phosphoryltriazolylalanine (1-pTza) and 3-phosphoryltriazolylalanine (3-pTza), analogues of 1-pHis and 3-pHis, respectively, which could be incorporated into peptides using conventional peptide synthesis strategies, and then deblocked to unshield the phosphate. Almost at the same time, Michael Webb’s group at Leeds University reported a very similar protocol for the synthesis of ethyl ester-protected Fmoc-3-pTza (McAllister et al., 2011). The Muir group went on to use Boc-3-pTza to synthesize a GAKR-3pTza-RKVLR peptide corresponding to the region around 3-pHis18 in histone H4, and then make a KLH-peptide conjugate for rabbit immunization; they showed that the resulting antibodies recognized pHis-containing H4 peptide and protein but not the unphosphorylated peptide (Kee et al., 2010). These seminal studies formally established that pHis antibodies could be generated.

This exciting advance prompted us to set up a collaboration with Mauger’s group. Taking a page from the playbook used by Bart Sefton’s group at the Salk Institute to generate anti-pTyr antibodies (Kamps, 1991), where they had used random Ala/Gly/pTyr-containing peptides as KLH-coupled antigens, we designed pTza peptide antigens with the 1- or 3-pTza analogue embedded in the middle of peptides with random Ala/Gly residues at four positions on either side, which could be coupled via an N-terminal Cys to KLH as a hapten carrier for immunization of rabbits (Fuhs et al., 2015). The rationale was to present the pTza analogues to the immune system in a peptide context, but one lacking significant immunogenic residues other than pTza itself. Our goal was to make sequence-independent anti-1-pHis and 3-pHis antibodies, analogous to the largely sequence-independent pTyr polyclonal and monoclonal antibodies that have been so useful. This venture turned out to be very successful, and we generated isoform specific anti-1-pHis and anti-3-pHis rabbit polyclonal sera. Subsequently, we used spleen cells from these rabbits to develop monoclonal antibodies (mAbs), obtaining three anti-1-pHis mAbs and four anti-3-pHis mAbs that with one exception are relatively sequence independent, pHis isoform selective and do not crossreact with pTyr (Fuhs et al., 2015). We have demonstrated that these mAbs can be used for immunoblotting, immunofluorescence staining of cells, immunohistochemical tissue staining, and for affinity purifying both pHis-containing proteins and pHis tryptic peptides for identification of pHis-containing proteins and histidine phosphorylation sites, respectively.

Several other groups have used 3-pTza or alternative stable 3-pHis analogues linked to hapten carrier proteins as immunogens with the goal of generating sequence-independent anti-3-pHis Abs (for a review of these efforts and original references see (Kalagiri and Hunter, 2021)). In some of the early attempts, the antibodies recognized 3-pHis proteins, but, as was the case with the 2G8 anti-pTyr mAb, also cross-reacted to a significant extent with pTyr proteins, which limited their utility. More recently, with newer 3-pHis analogues, the cross-reactivity problem has been significantly reduced. Although no structure of a pHis/pTyr cross-reacting mAb has been reported, the recent structures of the phosphoamino acid-binding pockets of the SC39–4 and SC56–2 3-pHis mAbs bound to a 3-pTza peptide and a 4G10 pTyr mAb CDR chimera bound to a pTyr peptide provide some insights into how the N³-phosphoimidazole moiety of 3-pHis and the O⁴-phenol ring of pTyr might be recognized by a single mAb (Kalagiri et al., 2021; Mou et al., 2018). Anti-pHis mAbs and polyclonal sera are now commercially available and have proved to be very valuable tools for the histidine phosphorylation community.

Histidine phosphorylation - present:

Histidine phosphorylation of proteins has been known for 60 years (Boyer et al. 1962), but progress in identifying new pHis-containing proteins and defining the sites and functional consequences of His phosphorylation has been slow, largely as a result of the experimental challenges inherent in studying such a labile protein modification. In consequence, our understanding of the functions of reversible His phosphorylation has lagged well behind that of canonical phosphorylations. However, recent technological advances are speeding progress. In particular, the availability of pHis-selective Abs and new methods for enriching and analyzing pHis proteins and peptides have begun to accelerate progress in the identification of pHis proteins and their His phosphorylation sites, enabling functional analysis of reversible His phosphorylation.

The pHis proteome:

Over the past few years, multiple approaches for identifying sites of phosphorylation in peptides with acid-labile phosphate linkages have been developed. In particular, several new methods for enrichment of pHis peptides have been reported, with enriched pHis peptides being analyzed by positive ion mode MS. For instance, a modified Fe²⁺-immobilized metal affinity chromatography (IMAC) phosphopeptide enrichment method run at pH 2.3 identified 246 pHis sites in a study of the E. coli phosphoproteome, representing ~12% of the total identified phosphosites (Potel et al., 2018). A separate survey of the E. coli pHis proteome using enrichment with a pan-pHis antibody raised against the pTze analogue of 3-pHis recovered 15 pHis sites (Oslund et al., 2014). In vertebrate systems, a study of zebrafish larvae using conventional TiO₂ phosphopeptide enrichment identified 68 pHis sites, which constituted 6.3% of the total identified phosphoproteome (Gao et al., 2019). The Eyers group developed a general enrichment method for acid-labile, non-canonical phosphopeptides, UPAX, which utilizes strong anion exchange chromatography at pH 6.8 to retain the full spectrum of canonical and non-canonical phosphosites. Using UPAX, they estimated that non-canonical phosphosites constituted ~30% of the total HeLa cell phosphoproteome, uncovering 83 pHis sites (Hardman et al., 2019). The Zhang group developed bis(zinc(II)-dipicolylamine)-derivatized silica microspheres (SiO₂@DpaZn) to enrich N-phosphopeptides under neutral conditions, with 19 of the 99 N-phosphosites identified in E. coli being pHis, and 611 of the 3384 N-phosphosites in HeLa cells being pHis (Hu et al., 2020). In a second HeLa cell study, pHis peptides were enriched with hydroxyapatite (HAP) chromatography in tandem with immunoaffinity purification on beads with a mix of immobilized 1- and 3-pHis mAbs, identifying 77 pHis peptides, including the known PGAM1 3-pHis11 site (Adam et al., 2019). Most recently, a strong cation exchange enrichment method was devised and applied to HeLa cells, identifying 563 pHis sites (Cui et al., 2021). To keep track of these burgeoning pHis datasets, an online pHis site database resource, HisPhosSite, was launched in 2021, which curates verified pHis sites reported in the literature and predicted pHis sites in a large number of prokaryotic and eukaryotic species (Zhao et al., 2021).

Taken together, these pHis site analyses suggest that the pHis proteome, especially in mammalian cells, might be quite large, and that the size of the entire “hidden” noncanonical phosphoproteome might rival that of the pTyr phosphoproteome. However, it should be noted that the published lists of pHis sites obtained by different investigators and approaches show relatively little overlap in phosphosite identity, even when the same human cell line is analyzed. In addition, very few of the authentic pHis sites, such as those in pHis enzyme intermediates, were identified in these surveys. This lack of congruence is a cause for some concern, and it remains possible that a fraction of the reported ”pHis” sites were in fact misidentified, even though strict false discovery limits were applied. One reason might be because the interpretation of pHis peptide MS spectra is complicated by bombardment-induced phosphate loss from the pHis residue, and its subsequent transfer to other residues in the peptide during positive ion mode MS analysis. For these reasons, further validation of any newly identified pHis site of interest will be essential, using site-directed mutagenesis, and ideally generation of site-specific pHis antibodies to prove the existence of a new pHis site. As an alternative approach for pHis site identification in peptides, negative ion mode MS analysis might be more reliable, because it is run under alkaline conditions where pHis is chemically very stable, although peptide fragmentation issues still need to be solved.

Histidine kinases:

The growing size of the pHis proteome in mammalian cells raises the question, what are the His kinases responsible for phosphorylating all these proteins? Eukaryotic His kinase activities were reported early on (Huebner and Matthews, 1985; Smith et al., 1973), but to date the only candidate eukaryotic His kinases are the NME family of nucleoside diphosphate (NDP) kinases (Besant et al., 2003; Wagner and Vu, 1995). These enzymes have a housekeeping function, where the phosphate from the 1-pHis enzyme intermediate is transferred onto an NDP molecule to regenerate the corresponding NTP. However, the NMEs can in principle also act as protein kinases by transferring the active site phosphate onto a His residue in a second protein. So far only the NME1 and NME2 members of this ten-member family have been shown to possess His kinase activity in vitro, and it is unclear whether the other members can also act as protein kinases, although most of them are known to autophosphorylate on their active site His. However, direct proof that NME1 and NME2 phosphorylate His residues in proteins in the cell has not been achieved. Also, based on the crystal structure of the pHis122 intermediate form of the Dictyostelium NDPK protein (Morera et al., 1995), it is not immediately obvious how the imidazole ring of a protein His residue could gain access to the phosphate linked to His122 for phosphate to be transferred. Nevertheless, a Pro96Ser mutation in NME1 has been reported to selectively reduce NME protein kinase activity relative to its NDP kinase activity (Freije et al., 1997; MacDonald et al., 1996), and, in principle, the P96S NME1 mutant could be used as a separation of function mutant in vivo. The combined knockout of NME1 and NME2 may be tolerated in specific mammalian cell types, but NME1/NME2 double knockout mice do not survive past birth (Boissan and Lacombe, 2011), presumably because of their essential housekeeping function in maintaining NTP levels. Currently, no selective NME1/2 inhibitors are available to test whether NME1/2 activity is required for intracellular phosphorylation of any particular protein on His. However, an NME1 activator compound has been reported, which could be useful in defining NME1/2 target proteins (Lee et al., 2018).

No other families of phosphotransfer proteins are obvious candidates to act as His kinases, although it remains possible that some of the uncharacterized conventional eukaryotic protein kinases might have His kinase activity. In addition, given that chemical phosphorylation of His sites in proteins and peptides is readily achieved in vitro under physiological conditions with phosphoramidate (O=P(OH)₂(NH₂)), another possibility is that some target His residues in proteins are phosphorylated in cells by small molecule phosphoramidates, such as creatine phosphate.

pHis phosphatases:

On the other side of the coin, if His phosphorylation is to serve a regulatory function, then protein-histidine phosphatase activities will be necessary to reverse the action of the His kinases. Currently, three pHis phosphatases (PHPs) that can dephosphorylate pHis are known (Jung et al., 2019): PHPT1, PGAM5, and LHPP, a haloacid dehalogenase (HAD) superfamily hydrolase that has been genetically linked to major depression (Neff et al., 2009). PHPT1 and PGAM5 dephosphorylate pHis in proteins in vitro, although the only pHis protein that PGAM5 is known to dephosphorylate is the pHis in NME2 itself (Panda et al., 2016). PGAM5 can also act as a pSer/pThr phosphatase, with dephosphorylation proceeding through a pHis intermediate. Moreover, although LHPP dephosphorylates the free pHis molecule, in the LHPP structure the active site is masked by a C2a-cap element, which normally restricts capped HAD family phosphatases of this sort to small molecule substrates, raising the question of how a pHis residue on a target pHis protein gains access to the LHPP active site (Gohla, 2019). There is a precedent, however, since some other C2a-capped HAD phosphatases, such as Eya, are known to act as protein phosphatases (Gohla, 2019).

One way to determine whether a putative PHP dephosphorylates pHis proteins in cells is to develop conditional PHP knockout cells and test if specific pHis proteins accumulate. Interestingly, Eyers’ group found that siRNA depletion of the PHPT1 pHis phosphatase in HeLa cells did not result in a gross increase in the number of pHis sites, either indicating the existence of other pHis phosphatases that have overlapping specificities, or that enzymatic dephosphorylation of pHis is not that important in HeLa cells (Hardman et al., 2019). In contrast, knockdown of PGAM5, a histidine acid phosphatase family member, in JURKAT T cells did increase the level of pHis NME2 (Panda et al., 2016). The reduced level of LHPP protein in hepatocellular carcinoma correlates with increased levels of pHis proteins (Hindupur et al., 2018), but whether LHPP acts directly or indirectly to promote dephosphorylation of these proteins in vivo is unknown. Ultimately, the importance of pHis phosphatases in the turnover of pHis on proteins in cells remains to be determined, and spontaneous hydrolysis of pHis in proteins, dependent on the sequence around the specific pHis site and a local acidic pH in the cell, might also play a significant role.

pHisphorylation as a regulatory mechanism:

More and more evidence that His phosphorylation serves to regulate protein function is being accumulated. For example, ion channel activity can be regulated through NME2-mediated phosphorylation. The calcium-activated potassium channel, KCa3.1, is activated in response to calcium influx in T cells through NME2-mediated phosphorylation of His358 in the KCa3.1 cytoplasmic tail, which is proposed to block coordination of an inhibitory Cu²⁺ ion by the four His358 residues in the channel tetramer (Panda et al., 2016; Srivastava et al., 2006; Srivastava et al., 2016). This process is negatively regulated by PHPT1 dephosphorylation of pH358 (Srivastava et al., 2008), and by PGAM5 dephosphorylation of pHis118 in NME2 (Panda et al., 2016). The TRPV5 calcium channel is activated by NME2 phosphorylation of His711 in the C-terminal tail, and inhibited by PHPT1 dephosphorylation of pHis711 (Cai et al., 2014). In contrast, based on reduced channel activity in Phtp1−/− mice, the TRPC4 cation channel may be negatively regulated by His phosphorylation. As another example, G-protein signaling can also be regulated by His phosphorylation (Srivastava, 2018). In this case, NME2 associates with the βγ subunits of heterotrimeric G-proteins, leading to phosphorylation of His266. In a subsequent step, pHis266 transfers its phosphate onto GDP to generate GTP, which then contributes to G protein activation and adenylyl cyclase activation (Cuello et al., 2003).

With regard to cellular processes that might be regulated by His phosphorylation, gene ontology analyses of mapped sites of His phosphorylation from several phosphoproteomic studies (see above) and proteins enriched with pHis mAbs (Fuhs et al., 2015) suggest that cell cycle progression, signal transduction, and RNA splicing/processing could all be targets. Gene ontology analysis of pHis sites in neuroblastoma cells implicated His phosphorylation in glycolysis, focal adhesion and cell migration, kinase signaling, and protein translation (Adam et al., 2020), and in HeLa cells macromolecular metabolism, microtubule function, and RNA processing and transcription were noted (Hu et al., 2020). Given the large number of pHis sites being identified, new functions for His phosphorylation will undoubtedly begin to emerge over the next few years as more pHis sites are validated and functionally characterized. In particular, roles for His phosphorylation in transcription may emerge, and the enigmatic function of His18 phosphorylation in histone H4 unveiled.

A possible role for His phosphorylation in cancer has recently attracted a lot of interest. The NME1/nm23 His kinase was first implicated in cancer nearly 35 years ago as a potential metastasis suppressor in melanoma and breast cancer (Steeg et al., 1988), with follow-on work from that group implicating the protein kinase activity of NME1 in inhibiting metastasis (Freije et al., 1997). However, subsequent studies have suggested a positive role for NME1 in prostate and ovarian cancer metastasis (Andolfo et al., 2011), and more needs to be done to elucidate the role of NME1 His kinase activity in metastasis. Identification of the key proteins whose phosphorylation by NME1 regulates metastasis will be an important step forward in this regard. More recently, several reports have suggested functions for His phosphorylation in promoting carcinogenesis. An increase in the levels of 1- and 3-pHis in proteins is observed in both hepatocellular carcinoma (HCC) (Hindupur et al., 2018) and in pediatric neuroblastoma (Adam et al., 2020), where the NME1 and NME2 genes are located in the Chr 17q amplicon. In both cancers NME1/2 protein levels are high compared to normal tissues, and in HCC the level of the LHPP PHP is significantly reduced. Identification of pHis proteins whose phosphorylation regulates their oncogenic function in these cancers should be a high priority.

Histidine phosphorylation - future:

What are the key questions that need to be answered to take His phosphorylation into the mainstream of protein phosphorylation? Having only two potential His kinases would seem to limit the possible repertoire of substrates, and efforts should be made to identify additional His kinases. For instance, is it possible that one or more of the conventional eukaryotic protein kinases (ePKs) has an atypical amino acid specificity? Of note here, Sgk196/POMK, a kinase that possesses all the ePK catalytic motifs, acts as a mannose kinase rather than a protein kinase (Yoshida-Moriguchi et al., 2013). Could mammalian proteins exist that switch the specificity of a Ser/Thr/Tyr protein kinase into a His kinase, as exemplified by the Salmonella SteE effector protein that switches GSK3 from a Ser/Thr kinase into a Tyr kinase (Panagi et al., 2020)? Finally, is it possible that some His sites are phosphorylated non-enzymatically by small molecule phosphate donors, such as IP7, mitochondrial acetyl phosphate, or creatine phosphate?

Although NME1 and NME2 reportedly have protein-histidine kinase activity in vitro, better evidence is needed to establish how NME1/2, which normally exist as dimers, can phosphorylate His residues in proteins, ideally through structures of a substrate protein bound to NME1 or NME2 in the presence of a nonhydrolyzable ATP analogue. In terms of defining new substrates for NME1/2 in vivo, proteomic approaches using affinity purification or proximity labeling with WT or kinase-dead mutant NME1/2, or the use of NME1/2 mutants with crosslinkable unnatural amino acids (UAAs) installed at key surface positions will identify interacting proteins that can be tested in vitro as NME1/2 substrates. Here, priority should be given to NME1/2 interacting proteins that have already been shown to contain pHis (Adam et al., 2020). In this regard, additional NME1/2 mutants that separate protein kinase from NDP kinase activity would be highly informative in determining how important NME1/2 are in His phosphorylation of cellular proteins, particularly if such separation of function mutations can be knocked into the endogenous NME1 or NME2 gene.

The role of the individual pHis phosphatases in dephosphorylation of pHis proteins should be explored in greater depth through creation of PHP gene knockouts and phosphatase-dead knock-in mutations, which can be used to evaluate the consequences of PHP loss for individual pHis phosphorylation events. The search for specific PHP-pHis substrate pairs will be aided by proteomic analysis of PHP-associated proteins, and biotinylation-based proximity proteomics, combined with pHis analysis of bound/biotin-tagged proteins to identify possible PHP substrates. An appropriately engineered catalytically-dead mutant form of the PHP may be useful as a substrate-trapping mutant, an approach that has been exploited successfully for identifying substrates of protein-tyrosine phosphatases. The extent to which PHPs might exhibit primary sequence specificity for pHis sites should also be explored.

The size of the pHis proteome needs to be more rigorously established. Some early estimates based on alkaline hydrolysis of proteins from ³²P-labeled cells put the levels of pHis as high as 5% of all protein-linked phosphate (Chen et al., 1977; Matthews, 1995), and recent whole cell ³¹P NMR data also estimate a high level of pHis (Makwana et al., 2020). Affinity purification of proteins from denatured lysates of FLAG-NME1 293 cells with a mixture of immobilized pHis mAbs resulted in identification of ~750 proteins; although there was no direct proof that all these proteins contained pHis, several known pHis-containing proteins were identified (Fuhs et al., 2015). As discussed above, MS analysis of pHis-containing tryptic peptides enriched by the newly developed methods has identified significant numbers of putative new pHis sites in mammalian cell lines and tissues, but to date catalogues of pHis sites obtained by different investigators and approaches show little overlap and contain very few of the known pHis sites. As analysis and validation methods are refined and pHis phosphosites validated, one expects that there will be better congruence between different pHis proteomic studies, and a truer picture of the size of the pHis proteome will emerge.

New canonical phosphorylation sites are routinely validated and functionally characterized by a defined series of experiments. These include making nonphosphorylatable and phosphomimetic mutations to demonstrate the site was correctly identified; these mutants can then be used to test the functional consequences of phosphorylation at that site. For pSer/pThr sites, Ala is commonly used as the nonphosphorylatable mutant, and Asp or Glu as phosphomimetic mutants, although neither Asp nor Glu has as strong a negative charge as phosphate, and they do not substitute for pSer or pThr to drive interactions with most phosphobinding domains. For pTyr, Phe is used as the nonphosphorylatable residue, but there is no good natural phosphomimetic amino acid. In part this problem has been overcome by the development of UAA analogues of pTyr that can be installed at a desired site in a protein in cells using expanded genetic code technology (Hoppmann et al., 2017). Finally, generation of sequence-specific phosphoantibodies to pSer/pThr/pTyr sites is commonly used as a further means to validate correct site identity and for subsequent functional analyses.

Functional validation of a site of His phosphorylation can follow a similar strategy, starting with generation of a nonphosphorylatable mutant to demonstrate loss of the pHis signal. For this purpose, Asn, which is isoteric with His, has been used, but lacks the positive charge required for a true His mimic; no natural amino acid is a good pHis mimetic. Development of a nonphosphorylatable positively-charged UAA analogue of His and nonhydrolysable UAA analogues of 1-pHis and 3-pHis would clearly be a valuable step forward. Sequence-specific 1/3-pHis antibodies have been made using nonhydrolyzable pHis analogue-containing peptides as antigens (Adam and Hunter, 2018; Kee et al., 2010), and such antibodies will be useful for assessing stimulus-induced changes in His phosphorylation of the target protein, and the subcellular localization of its pHis isoform. In this regard, one of the existing 3-pHis mAbs, SC44, exhibits a pseudo sequence specificity, preferring Gly-3pHis-Ala motifs, such as those found in the active sites of ACLY and SCSα (Kalagiri et al., 2021).

Another conundrum is how target histidine residues in proteins are selected by His kinases, and what dictates whether 1- or 3-pHis is formed at a particular site. Some conventional protein kinases initially recognize their substrates through short sequence motifs, known as docking motifs that interact with a docking site on the catalytic domain that is independent of the catalytic site, and then, once bound, the phosphorylation site, which is usually close in linear sequence to the docking motif, is selected. Whether NME1/2 use the docking site principle for recognizing their substrates is unknown but may emerge from further studies of NME1/2-interacting proteins and their sites of His phosphorylation. Likewise, structural analysis of His kinase/substrate interaction will be needed to understand the spatial constraints that dictate whether the target His is phosphorylated on the N1 or N3 position. Most conventional protein kinases exhibit primary sequence selectivity, with specific amino acids or charges being preferred at different positions up to five residues N- or C-terminal to the target hydroxyamino acid. Such sequence preferences have been deduced using degenerate peptide library approaches and sequence logo analysis of pSer/pThr/pTyr phosphosites phosphorylated by the protein kinase in question either in vitro or in vivo. However, logo sequence analysis of recent pHis proteomic datasets has not revealed any strong primary sequence preference, although there is often a preponderance of Leu and Arg/Lys residues in the vicinity of the target His (Adam et al., 2019; Hardman et al., 2019; Hu et al., 2020). Whether these apparent amino acid selectivities reflect a true phosphorylation site preference, or whether there is a more trivial explanation is unclear. For instance, the enrichment method itself might select pHis peptides of this sort. Alternatively, pHis peptides with this composition might be better protected from hydrolysis, particularly when they are exposed to the strongly acidic conditions required to ionize the peptides prior to injection into the MS instrument.

Another question is whether there are secondary structure requirements for His phosphorylation, and how this might affect whether the N1 or N3 position is phosphorylated. ePKs generally phosphorylate target residues in locally unstructured regions of proteins requiring an extended peptide backbone to bind into the kinase active site. pHis residues in deposited crystal structures occur in helices, loops and β-strands (see (Kalagiri and Hunter, 2021)). However, nearly all of these structures are pHis enzyme intermediates, where there are structural constraints on the orientation of the pHis residue dictated by the nature of the active site, and it seems unlikely that such pHis enzyme structures will reflect those of pHis in reversibly regulated His kinase substrates. For instance, the pHis18 site in the N-tail of histone H4 is probably unstructured. His358 in the C-terminal extension of the KCa3.1 calcium-activated potassium channel was predicted to be in a four-helix bundle (Srivastava et al., 2016), although this conclusion needs to be reevaluated in light of the more recent cryo-EM structure of the full length SK4/KCa3.1 channel (Lee and Mackinnon, 2018). Clearly, additional structures of pHis-containing proteins will be highly informative and teach us more about how His phosphorylation regulates protein function.

Sequence-specific phosphobinding domains exist for pSer, pThr and pTyr residues and these have important functions in phosphorylation-mediated signal transmission. Could pHis signals be transmitted similarly through pHis-binding domains? Given that pTyr and 3-pHis have structural similarities, and that some 3-pHis Abs cross-recognize pTyr and vice versa, the existence of pHis-binding protein modules seems plausible, and further search for pHis-binding proteins using the stable pHis analogues would be worthwhile.

Technology development will undoubtedly play an important role in accelerating our understanding of His phosphorylation. Using protein engineering to improve the selectivity and affinity of pHis mAbs will be a useful advance, and the generation of site-specific pHis antibodies for functionally important pHis sites will be very valuable. The development of more definitive MS methods for identifying pHis sites unequivocally, and particularly negative ion mode MS, should be an important goal. Genetic tools for manipulating pHis phosphorylation in vivo, including the use of UAA analogues of unphosphorylated His and 1/3-pHis will be key to understanding the functions of individual His phosphorylation sites. Another approach for perturbing His phosphorylation processes in living cells would be to use Inducible expression of GFP-tagged scFv versions of pHis mAbs, to localize pHis proteins in the cell, stabilize pHis sites, and to disrupt the function of His phosphorylation. The development of selective small molecule inhibitors of His kinases and pHis phosphatases would also provide powerful tools. However, whether it will be possible to design a selective inhibitor of NME1/2 His kinase activity that does not affect their housekeeping NDPK function is unclear, and development of an allosteric inhibitor that prevents protein substrate binding might be the best option.

Ultimately, we need to discover how reversible His phosphorylation regulates protein function, and what sort of biological processes are regulated. Given the large number of His phosphorylation sites, prioritization of which sites to study will be essential. Based on the GO categories for recently identified pHis sites discussed above, investigation of pHis proteins involved in RNA metabolism seems like a potentially profitable area, because it is easy to imagine how phosphorylation of key His residues in RNA-binding domains could regulate target RNA binding, and also disrupt the binding of divalent cations important for RNA function (Fuhs and Hunter, 2017). pHis proteins involved in cell cycle progression, cancer and affective disorders are other areas worthy of consideration. From my perspective, some of the most important open questions concerning the function of His phosphorylation are listed in Box 1.

Box 1. Open questions about the function of His phosphorylation.

• What types of cellular functions are regulated by His phosphorylation?

• Given its chemical instability, is pHis particularly important for short-term responses?

• Are all His phosphosites functionally significant, or are some pHis sites adventitious?

• Are proteins selectively phosphorylated at the 1- or 3-position of the target His, and, if they are, what dictates the choice? His538 in the KCa3.1 channel is exclusively phosphorylated on the N-3 position.

• How does His phosphorylation regulate protein activity? Is there a generalized principle or are there multiple mechanisms depending on the particular pHis site, as is the case for canonical phosphorylation events?

  • Could pHis act through local charge effects on proteins? The charge on a His residue goes from +1 to −2 upon addition of a single phosphate.
  • Does His phosphorylation trigger regulatory allosteric changes or induce protein-protein interactions (PPIs)?
  • Does His phosphorylation act by regulating divalent metal ion binding through phosphorylation of His coordination residues?

• Are there pHis-specific binding domains (c.f. SH2) that can be used to transmit pHis signals and mediate PPIs?

• Where are pHis proteins localized in cells and could His phosphorylation lead to changes in localization?

• Could local differences in intracellular pH dictate the stability of pHis residues?

• Under what cell conditions are the levels of pHis in proteins changed – growth factors, attachment, intracellular pH, etc.?

• Is there crosstalk between pHis and other phosphoamino acids or other PTMs at the protein level?

• To what extent does aberrant His phosphorylation play a role in disease?

• Can His phosphorylation networks be targeted for therapeutic purposes in diseases such as cancer and affective disorders?

And going beyond His phosphorylation, what we have learned about His phosphorylation will undoubtedly be very valuable in studies of other noncanonical protein phosphorylations, and particularly in investigating the functions of the other two phosphoramidate derivatives, pLys and pArg, which are being identified in increasing numbers in phosphoproteomic studies.

Coda:

Despite the rapid progress in the 25 years since Benjamin Lewin started Molecular Cell as a “molecular version of Cell” in 1997, a great deal remains to be done to understand how protein phosphorylation networks are integrated and utilized. In my view, some of the outstanding questions we need to resolve are shown in Box 2. Over the next decade, determining the functions of individual phosphosites will continue to be a major undertaking, and then, going beyond protein phosphorylation per se, much needs to be done to understand the intricate crosstalk between phosphorylation and the myriad of other post-translational modifications, the topic of my review for the 10^th anniversary of Molecular Cell in 2007 (Hunter, 2007).

Box 2. Open questions about protein phosphorylation.

• How important are noncanonical phosphorylations in cellular regulation and organismic physiology?

• What fraction of the thousands of identified phosphosites are adventitious or silent, but yet are tolerated by the cell?

• How important is the stoichiometry of phosphorylation at specific sites?

• How fast do phosphates turn over at individual phosphosites on proteins in the cell, and can the rate of phosphate turnover be used informationally, for instance in oscillatory circuits?

• How many sites are phosphorylated on a single protein molecule at any one time and are specific combinations of phosphates functionally important, and from a technical viewpoint how can one determine which sites are occupied at the single protein molecule level (e.g., nanopore-based single protein molecule sequencing)?

• How are the outputs of multiple occupied phosphosites on a single protein molecule coordinated and integrated?

• Do new principles of regulation by protein phosphorylation await discovery? For instance, could phosphorylation and dephosphorylation of intrinsically disordered regions (IDRs) in proteins play a key role in the formation and disassembly of biomolecular condensates?

Supplementary Material

1

Figure S1: Excerpts taken from pages of Martin Broome’s lab notebooks from December 1989 and March 1990 describing his futile efforts to generate phosphohistidine antibodies. The key comments are:

May. #3. Do ELISAs on pAb (polyclonal) serum against P-His. Failure!!! 5/11: B) 3: Basically, no evidence for anti-PHis. Waaaah!